66 lines
3.0 KiB
Markdown
66 lines
3.0 KiB
Markdown
|
---
|
||
|
|
license: apache-2.0
|
||
|
|
datasets:
|
||
|
|
- wikipedia
|
||
|
|
- mc4
|
||
|
|
- cc100
|
||
|
|
- oscar
|
||
|
|
language:
|
||
|
|
- ja
|
||
|
|
---
|
||
|
|
# japanese-large-lm-3.6b
|
||
|
|
|
||
|
|
This repository provides a 3.6B parameters Japanese language model, trained by [LINE Corporation](https://linecorp.com/ja/).
|
||
|
|
|
||
|
|
[Tech Blog](https://engineering.linecorp.com/ja/blog/3.6-billion-parameter-japanese-language-model) explains details.
|
||
|
|
|
||
|
|
## How to use
|
||
|
|
|
||
|
|
```
|
||
|
|
import torch
|
||
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline, set_seed
|
||
|
|
|
||
|
|
model = AutoModelForCausalLM.from_pretrained("line-corporation/japanese-large-lm-3.6b", torch_dtype=torch.float16)
|
||
|
|
tokenizer = AutoTokenizer.from_pretrained("line-corporation/japanese-large-lm-3.6b", use_fast=False)
|
||
|
|
generator = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)
|
||
|
|
set_seed(101)
|
||
|
|
|
||
|
|
text = generator(
|
||
|
|
"おはようございます、今日の天気は",
|
||
|
|
max_length=30,
|
||
|
|
do_sample=True,
|
||
|
|
pad_token_id=tokenizer.pad_token_id,
|
||
|
|
num_return_sequences=5,
|
||
|
|
)
|
||
|
|
|
||
|
|
for t in text:
|
||
|
|
print(t)
|
||
|
|
|
||
|
|
# 下記は生成される出力の例
|
||
|
|
# [{'generated_text': 'おはようございます、今日の天気は雨模様ですね。梅雨のこの時期の 朝は洗濯物が乾きにくいなど、主婦にとっては悩みどころですね。 では、'},
|
||
|
|
# {'generated_text': 'おはようございます、今日の天気は晴れ。 気温は8°C位です。 朝晩は結構冷え込むようになりました。 寒くなってくると、...'},
|
||
|
|
# {'generated_text': 'おはようございます、今日の天気は曇りです。 朝起きたら雪が軽く積もっていた。 寒さもそれほどでもありません。 日中は晴れるみたいですね。'},
|
||
|
|
# {'generated_text': 'おはようございます、今日の天気は☁のち☀です。 朝の気温5°C、日中も21°Cと 暖かい予報です'},
|
||
|
|
# {'generated_text': 'おはようございます、今日の天気は晴天ですが涼しい1日です、気温は午後になり低くなり25°Cくらい、風も強いようですので、'}]
|
||
|
|
```
|
||
|
|
|
||
|
|
## Model architecture
|
||
|
|
| Model | Vocab size | Architecture | Position type | Layers | Hidden dim | Attention heads |
|
||
|
|
| :---: | :--------: | :----------- | :-----------: | :----: | :--------: | :-------------: |
|
||
|
|
| 1.7B | 51200 | GPT2 | Absolute | 24 | 2304 | 24 |
|
||
|
|
| 3.6B | 51200 | GPTNeoX | RoPE | 30 | 3072 | 32 |
|
||
|
|
|
||
|
|
## Training Corpus
|
||
|
|
Our training corpus consists of the Japanese portions of publicly available corpus such as C4, CC-100, and Oscar.
|
||
|
|
We also incorporated the Web texts crawled by in-house system.
|
||
|
|
The total size of our training corpus is about 650 GB.
|
||
|
|
The trained model achieves 7.50 perplexity on the internal validation sets of Japanese C4.
|
||
|
|
|
||
|
|
## Tokenization
|
||
|
|
We use a sentencepiece tokenizer with a unigram language model and byte-fallback.
|
||
|
|
We **do not** apply pre-tokenization with Japanese tokenizer.
|
||
|
|
Thus, a user may directly feed raw sentences into the tokenizer.
|
||
|
|
|
||
|
|
|
||
|
|
## License
|
||
|
|
[Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0)
|